Hi Chris,
The base suckout current that is needed for BJTs is often surprizingly high for many, even for reasonably fast BJT's. I have a whole section on the math for that in my book.
The suckout current needed is dependent on the output current slew rate, in A/us. Obviously, this gets worse driving lower load impedances at higher frequencies. In some cases, it gets even worse when the power transistor is experiencing ft droop at high output current and simultaneous high output current slew rate. An example of this is driving a square wave close to full power into a resistive load. Things can get even worse under some conditions when driving reactive loads.
I don't think 30 watts of quiescent power for a beefy high-quality 100W amplifier is unreasonable, especially when you look at how much the amplifier dissipates when driving 1/3 power into an 8-ohm load. Also bear in mind that while my example MOSFET design had 150 mA idle bias, an optimally biased BJT design with 0.22-ohm emitter resistors will idle at about 120 mA.
For high sound quality and freedom from crossover distortion and transient under-bias, you want to be able to run a healthy amount of quiescent bias current without getting into a lot of gm doubling in the BJT case (MOSFETs virtually never can be biased into gm doubling). This also gives you a bigger class A region of operation.
Its just the old heat vs. sound quality tradeoff. If you want to run a BJT stage at only 60 mA bias and with 0.47 ohm emitter resistors, you certainly can, but performance will not be great, in my opinion.
Cheers,
Bob
Hi Tad,
Yes, indeed. I won't jump the gun right now, but after the book is out you'll know who it was. He is someone whom I truly trust and admire. Someone who is brilliant, yet truly generous with his time. Someone to whom I am truly indebted. I am blessed with many good friends.
Cheers,
Bob
Hi Bob,
Yes, pretty much what I was thinking of. Still, with 0.1 to 0.22 ohm emitter resistors, I've seen the "happy spot" for bias sitting around 16 mA per pair. It's a good thing I've already ordered your book. Looking at the interest here, they may sell out. It would make an excellent text for teaching in university, along with some notable others.
The On-Semi MJW0281A and MJW0302A pairs are about the nicest BJTs I have seen yet for performance yet. Very closely matched and were (before they were discontinued) worthy of a look. They are now used in a Thermal Track part, but being a lower power rating I feel many will overlook them. Still, the newer On-Semi parts are better than what we all grew up with. It took a long time to catch up to the better Japanese parts. Now, if they can just come up with some low and mid power devices of the same quality!
-Chris
Yes, pretty much what I was thinking of. Still, with 0.1 to 0.22 ohm emitter resistors, I've seen the "happy spot" for bias sitting around 16 mA per pair. It's a good thing I've already ordered your book. Looking at the interest here, they may sell out. It would make an excellent text for teaching in university, along with some notable others.
The On-Semi MJW0281A and MJW0302A pairs are about the nicest BJTs I have seen yet for performance yet. Very closely matched and were (before they were discontinued) worthy of a look. They are now used in a Thermal Track part, but being a lower power rating I feel many will overlook them. Still, the newer On-Semi parts are better than what we all grew up with. It took a long time to catch up to the better Japanese parts. Now, if they can just come up with some low and mid power devices of the same quality!
That's true enough for many folks, but I can't help but think of all the people who leave their stuff on all the time (don't!). All that heat must be removed by air conditioning, or the winter chill (and then it's fine). I do have a personal bias <groan> against things running hot unless it's absolutely necessary. In a case like that, current levels accepted, but it still bothers me. No, I'm not a green energy guy. I like me big trucks and real station wagons!
On that, we both agree!
-Chris
Linear Audio - It's here!
Jan, Congratulations!!!
I just got home and my copy of Linear Audio was waiting for me in my mail box. Wow, what a wonderful piece of work! This is a book! I am truly proud to be an author in your inaugural issue and wish you the very, very best with the future of Linear Audio.
You have made a valuable contribution to the health and vibrance of audio DIY.
Best regards,
Bob
Jan, Congratulations!!!
I just got home and my copy of Linear Audio was waiting for me in my mail box. Wow, what a wonderful piece of work! This is a book! I am truly proud to be an author in your inaugural issue and wish you the very, very best with the future of Linear Audio.
You have made a valuable contribution to the health and vibrance of audio DIY.
Best regards,
Bob
Hi Bob,
Yes, pretty much what I was thinking of. Still, with 0.1 to 0.22 ohm emitter resistors, I've seen the "happy spot" for bias sitting around 16 mA per pair. It's a good thing I've already ordered your book. Looking at the interest here, they may sell out. It would make an excellent text for teaching in university, along with some notable others.
-Chris
Hi Chris,
You make an interesting observation, which I have seen also, and which I show a curve of in my book in the section on thermal bias stability - namely that the apparent optimal quiesecent bias current for a BJT class AB output stage is often less than the theoretical optimum that places 26 mV across each emitter resistor.
There are a couple of potential contributors to this phenomena. One is RB of the output transistors plus the resistance of base stopper resistors, if used. These result in an effective ohmic resistance at the emitter that actually counts as part of the external emitter resistance insofar as the optimum bias current calculation goes.
Consider an output transistor with RB=5 ohms and a 5-ohm gate stopper resistor, where the current gain of the output transistor is 50. This causes an ohmic component of emitter resistance to appear seen looking into the emitter of 0.2 ohms. This resistance effectively adds to the total emitter resistance. If the amp had 0.22 ohm external emitter resistors, it "thinks" it has 0.42 ohm emitter resistors and wants to be biased such that 26 mV is placed across 0.42 ohms, NOT 0.22 ohms. This puts only 13.6 mV across the external 0.22 ohm resistors, for an "optimum" quiesecent current of 62 mA instead of 118 mA. Bear in mind the example here may be a bit extreme. The lesson here is that we cannot blindly bias for 26 mV across RE and expect to meet the Oliver criteria for optimum class AB bias when using real transistors and possibly gate stopper resistors.
Unfortunately, the story does not end here. There is at least one other fly in the ointment. That is the operating junction temperature of the transistors. The junctions run hotter when some signal is going through the amplifier into a load, even at fairly low power levels. For example, if you operate the amplifier at a low power level where you think crossover distortion is maximum, and adjust bias for lowest crossover distortion, then take signal away for a several seconds, you will find that the voltage across the emitter resistors is smaller than 26 mV. This is because the junction temperature is different, even seconds after signal is removed.
Conversely, if you bias the amplifier to 26 mV after it has warmed up and stabilized under no-power conditions, you will find that under signal conditions the amplifier is over-biased because the junctions of the transistors run hotter when signal is present (even only a couple of watts). This phenomena is there even if you have a huge heat sink. I have a whole chapter devoted to this kind of stuff. Fortunately, the use of ThermalTrak transistors greatly mitigates this, and I have some curves that illustrate it.
Cheers,
Bob
Have a look at Roender's set up procedure. He uses ~18 to 19mVre (across the external Re) as the optimum bias because he has found precisely this external/internal resistor interaction.
Many others have made similar comment.
As for 26mV as the total Vre (internal + external), that seems to only apply when Tj=25degC, according to the literature I have read through.
It appears that many believe this 26mVre must be adjusted downwards as junction temperature rises. What law should this variation in optimum bias voltage follow to maintain that minimum distortion with varying Tj?
Is that in the book?
There may be a good argument that Tj varies less when the transistors are run hot due to high optimum bias current.
That low (external) Re combined with the closely coupled diodes may well be why Roender's implementation has impressed so many.
Many others have made similar comment.
As for 26mV as the total Vre (internal + external), that seems to only apply when Tj=25degC, according to the literature I have read through.
It appears that many believe this 26mVre must be adjusted downwards as junction temperature rises. What law should this variation in optimum bias voltage follow to maintain that minimum distortion with varying Tj?
Is that in the book?
There may be a good argument that Tj varies less when the transistors are run hot due to high optimum bias current.
That low (external) Re combined with the closely coupled diodes may well be why Roender's implementation has impressed so many.
Last edited:
Hi Andrew,
Actually, I posted this in the original thread where the bias currents were recommended to be higher. The current I posted was only valid for the MJW devices of course and other devices (properly) required higher currents.
Hi Bob,
Thank you. I've seen this over the course of many, many years in service. Manufacturers recommendations err on the side of too high (before it was popular for a warm heat sink to be good). The thinking was that it did avoid crossover distortion and the resulting distortion seemed to be more benign to the listener. It didn't help that many designs didn't really have good control over the bias current either. In fact, the mid to late 70s Marantz amplifiers often had closer control and ended up being slightly negative for temperature coefficient. They seemed to be more stable and sound better as long as you looked after the basics when you serviced it. Purely observation here. The class "B" designs typically had a edginess to their sound, including McIntosh. Even the early tube product was this was, the 240 sounding the best in my opinion. MC30 and 40 weren't something I would own for very long (sold my MC40s soon after doing a rebuild).
Your observations regarding internal effective emitter resistance are also right on. 0.47 ohm emitter resistors seem to preclude the ability to bias to the correct point from what I've seen. Not a happy setup. Still, for some designs a 0.47 R resistance is the only way they keep the bias current in check. Sloppy work there, but more common than people want to think.
The newer On-Semi perforated emitter designs really stepped the quality of their power transistors up in a major way. I feel a connection to Toshiba there. I'm very interested to see what you have found with your examinations of these. In particular, the MJW0281A and MJW0302A took me completely by surprise. Not only was the beta distribution tight, but the beta between NPN and PNP parts were very close. After testing the first ones, I bought more and confirmed this same trend. Replacing some outputs in existing prototypes showed the lower bias currents required to get a distortion null. I don't really know what all is going on in these, but the distortion numbers without feedback were very low in comparison to other Japanese parts, and On-Semi's earlier designs. The sound quality also seems to confirm this from a casual listening point of view of prototypes with traditional feedback. Hearing these were discontinued was not a great event as far as I'm concerned, but at least these did show up as a thermal track part. As for hot spots on the die, I suspect that these transistors have far better current distribution than the normal perforated emitter designs previously done. That would explain the stability.
Bob, you've given me even more reason to watch for this book. You may not see me for a bit as I read through it! I'm extremely happy that you did examine the new On Semi parts.
-Chris
Yes.
Actually, I posted this in the original thread where the bias currents were recommended to be higher. The current I posted was only valid for the MJW devices of course and other devices (properly) required higher currents.
Hi Bob,
Thank you. I've seen this over the course of many, many years in service. Manufacturers recommendations err on the side of too high (before it was popular for a warm heat sink to be good). The thinking was that it did avoid crossover distortion and the resulting distortion seemed to be more benign to the listener. It didn't help that many designs didn't really have good control over the bias current either. In fact, the mid to late 70s Marantz amplifiers often had closer control and ended up being slightly negative for temperature coefficient. They seemed to be more stable and sound better as long as you looked after the basics when you serviced it. Purely observation here. The class "B" designs typically had a edginess to their sound, including McIntosh. Even the early tube product was this was, the 240 sounding the best in my opinion. MC30 and 40 weren't something I would own for very long (sold my MC40s soon after doing a rebuild).
Your observations regarding internal effective emitter resistance are also right on. 0.47 ohm emitter resistors seem to preclude the ability to bias to the correct point from what I've seen. Not a happy setup. Still, for some designs a 0.47 R resistance is the only way they keep the bias current in check. Sloppy work there, but more common than people want to think.
The newer On-Semi perforated emitter designs really stepped the quality of their power transistors up in a major way. I feel a connection to Toshiba there. I'm very interested to see what you have found with your examinations of these. In particular, the MJW0281A and MJW0302A took me completely by surprise. Not only was the beta distribution tight, but the beta between NPN and PNP parts were very close. After testing the first ones, I bought more and confirmed this same trend. Replacing some outputs in existing prototypes showed the lower bias currents required to get a distortion null. I don't really know what all is going on in these, but the distortion numbers without feedback were very low in comparison to other Japanese parts, and On-Semi's earlier designs. The sound quality also seems to confirm this from a casual listening point of view of prototypes with traditional feedback. Hearing these were discontinued was not a great event as far as I'm concerned, but at least these did show up as a thermal track part. As for hot spots on the die, I suspect that these transistors have far better current distribution than the normal perforated emitter designs previously done. That would explain the stability.
Bob, you've given me even more reason to watch for this book. You may not see me for a bit as I read through it! I'm extremely happy that you did examine the new On Semi parts.
-Chris
Hi Andrew,
The idealized solution with ideal transistors says that Vq = Vt = KT/q, which is about 26 mV at room temperature, where Vq is the voltage across each emitter resistor. If this were the only thing at work, or dominated, one would argue that Vq should actually increase in proportion to absolute temperature of the junction. Of course, this effect is not the only effect at work.
Even if changes in Vt were the only thing at work, such absolute-temperature-dependent phenomena must be put into perspective. Suppose a transistor dissipates 5W at idle, the heat sink is at 40C at idle, and thermal resistance from junction to case through insulator to heat sink is 1.3 C/W. The junction will be at about 47C.
Now suppose under program conditions the average transistor power dissipation has risen to 30W and the heat sink is big and has a lot of thermal mass and has not moved. The junction will now be at about 79C.
This means that the junction temperature has risen from 320K to 352K, or an increase of 10%. We have much bigger fish to fry and much bigger problems to confront than a mere 10% error introduced by changes in Vt. Those other problems I mentioned earlier are much bigger.
BTW, if you have to err in bias, it is definitely better to err on the over-bias side as long as your amplifier has good thermal stability so that it will never run away.
Cheers,
Bob
The ThermalTrak parts are great, and I was very pleased with my experimental results. In one case I built up an amplifier and closed the feedback loop ahead of the output stage. I set the amplifeir up so that I could bias it with the ThermalTrak sensing diodes or with a conventional heat sink tracking approach. The difference in thermal bias stability after a thermal step was really dramatic. I also had a lot of fun coming up with variations on the ThermalTrak bias spreader schemes.
Cheers,
Bob
VT, auto bias
Hi Bob,
Several auto bias circuits, discrete or even integrated (LT1166) rely on this equation: Vt = KT/q. IOW, the bias current is proportional to the abs. temp. (of said circuit, thus not the temp. of output devices). In light of the preceding considerations, should this be regarded as asset or as flaw?
Cheers,
E.
Hi Bob,
Several auto bias circuits, discrete or even integrated (LT1166) rely on this equation: Vt = KT/q. IOW, the bias current is proportional to the abs. temp. (of said circuit, thus not the temp. of output devices). In light of the preceding considerations, should this be regarded as asset or as flaw?
Cheers,
E.
Hi Bob,
Try using the diodes in series with the emitters of the input transistors in a diamond buffer arrangement. I am using these in an arrangement that also includes a CFP output stage with another ThermalTrak diode to bias those transistors. There are really interesting things that these diode arrangements allow you to do that wasn't possible before hand. Just because these devices are in the output parts doesn't mean they can't be used further back in the chain.
A few years ago, the folks at Denon used little glass signal diodes running between the emitter - collector leads of the driver transistors. That was brilliant thinking for the time. Someone over there is obviously up on their knowledge. I hadn't thought of this until I saw that arrangement, then it hit me like a ton of bricks. Inspired work - whoever drew that up.
Hi Arthur,
Not wanting to head off Bob's possible reply to your question, but there are many factors that affect the sound of an amplifier. There is really no "best amplifier circuit" either. The end result relies heavily on the physical execution of the design too.
I don't really know if there is an answer to your question, and then it's part personal taste as well.
Hi Edmond,
The LT1166 didn't seem to perform well for me. A commercial design I had in did not sound very good either. Did you have any success with it?
-Chris
Try using the diodes in series with the emitters of the input transistors in a diamond buffer arrangement. I am using these in an arrangement that also includes a CFP output stage with another ThermalTrak diode to bias those transistors. There are really interesting things that these diode arrangements allow you to do that wasn't possible before hand. Just because these devices are in the output parts doesn't mean they can't be used further back in the chain.
A few years ago, the folks at Denon used little glass signal diodes running between the emitter - collector leads of the driver transistors. That was brilliant thinking for the time. Someone over there is obviously up on their knowledge. I hadn't thought of this until I saw that arrangement, then it hit me like a ton of bricks. Inspired work - whoever drew that up.
Hi Arthur,
Not wanting to head off Bob's possible reply to your question, but there are many factors that affect the sound of an amplifier. There is really no "best amplifier circuit" either. The end result relies heavily on the physical execution of the design too.
I don't really know if there is an answer to your question, and then it's part personal taste as well.
Hi Edmond,
The LT1166 didn't seem to perform well for me. A commercial design I had in did not sound very good either. Did you have any success with it?
-Chris
